3D-BBDA: Backpropagation-Free Black-Box Domain Adaptation for Robust 3D Point Cloud Recognition

This research presents 3D-BBDA, a framework for robust 3D point cloud recognition that adapts source-trained models to unlabeled target data without requiring access to the source model's internal parameters or data. The method uses pseudo-label guided prototype-based alignment and a weighted prediction module to fuse logits, while employing a Gaussian-perturbed prototype prior to stabilize memory during incomplete coverage. Experiments on four corrupted 3D benchmarks show consistent improvements over baselines; specifically, on ModelNet-40C, 3D-BBDA achieves mean errors of 32.08% (PointNet) and 29.62% (GDANet), outperforming Source-only (35.88% / 34.03%) and BFTT3D (33.78% / 32.36%). The approach requires only a single forward pass with no gradient updates, making it suitable for real-time, privacy-preserving deployment in transferable 3D recognition pipelines.